Output Explorer

Every prompt in the paper, and what each model wrote back.

Answer a question about a scanned page from its OCR text. Scored by exact match and ANLS against the accepted answers.

13 of 5,330 prompts

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageqjvj0224_2.png

Which is the least critical issue for live rep support?

OCR text of the page · 1,263 characters; the scanned image itself is not published
July 31, 2003
Consumer Relations Efficiency Team
Team:
Nancy Bowland
Sharonda McMurray
Veronica Walton
Gary Hicks
Nancy Montgomery
Yvette Willard
Todd Holbrook
Brice O'Brien
Cindi Hunter
Donna Walkup
Objective:
Balance cost efficiency with quality customer service
Our Process:
Examined Consumer Relations cost structure:
West Interactive - Automation
- AT&T
Bellomy - Live Rep
- Brand Chargebacks
YA - Live Rep
- Vertis
Ghostwriters
Reasons for Consumer Contact:
%
Request Catalog/Order Form
55%
Order Status
19%
Mailing List Request (add, change, update add, target change)
14%
Promotional Questions (lists, seals/c-notes questions, terms, Web issues, etc.)
5%
Product Issues
4%
Non-receipt DM, conversion, fulfillment
1%
ELP Issues - card request, statement issues
1%
Other
1%
Conducted a zero-based planning exercise - Hierarchy of Importance for Live Rep Support
Non-Critical
Retailer calls/
other issues
Live Rep
Catalog/order form requests
Access
Promotional program issues
order status, complaints)
Brand product/packaging changes
Mailing list additions/Database updates
Critical
Product Quality/Liability Issues
. Examined 4 levels of service options ranging from $1.IMM to $6. IMM.
52975 1217
Source: https://www.industrydocuments.ucsf.edu/docs/qjvj0224
System prompt · identical for every setup
Answer the question using only the OCR text from a single document page. Return only the answer, with no explanation. Preserve the answer wording from the OCR text when possible.
Expected answer
Retailer calls/ Other issuesRetailer calls/Other issues
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.68

Non-Critical Retailer calls/ other issues

41 characters10 tokens

Aux 2015

Wrong

The least critical issue for live rep support is the non-receipt of a customer's order.

91 characters21 tokens

PiT-FT 2015

Wrong

<|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|>

120 characters64 tokens

ChronoGPT 2015

Wrong
Product/packaging changes
Product/packaging changes
Product/packaging changes

The same 26-character fragment repeats 10 times until the token limit. Showing the first three.

277 characters64 tokens